ANSI-C program: frost.c
NAME
frost - Minimum mean square error filter (Frost et al.,
1982).
SYNOPSIS
frost <pwr1> <pwr1_frost> <width> [fx]
[sx] [power]
<pwr1> | input intensity image (float values) |
<pwr1_frost> | output intensity image (float values) |
<width> | number of samples/row |
[fx] | filter window size (3,5,7,...) |
[sx] | window size used for statistics (3,5,7,...) |
[power] | power applied to weighting coefficients alpha, (default:1.0 corresponds to original filter), power > 1.0 leads to an improved edge preservation. |
EXAMPLE
frost 1352_1610.pwr1 1352_1610.pwr1_frost 2500 3 7
1.0
frost 1352_1610.pwr1 1352_1610.pwr1_frost 2500 3 7 2.0
The power 2.0 applied to the weighting coefficients alpha results in a sharper filter, i.e. edges are even better maintained.INSTALLATION
Source code frost.c in ./src. For compilation adjust and use
Makefile: Executable version frost in ../bin
AVAILABILITY
Uses ISP type definition file typedef_ISP.h.
DESCRIPTION
The Minimum Meas Square Error Filter (MMSE) as described by Frost
et al., (A Model for Radar Images and its Applications to digital
filtering of multiplicative noise, IEEE Trans. Pattern Analysis
and Machine Intelligence Vol. PAMI-4, No. 2, pp. 157-165, 1982)
is an adaptive filter for SAR data. The locally used estimator is
deteremined based on the statistics of a window of userdefined
size fx. The size of the window used for the filtering, sx, may
deviate from the window used to determine the local statistics.
For more homogeneous areas stronger filtering is applied, for
more heterogeneous areas reduced filtering is applied. As a
result the filter preserves strong single scatterers and
edges.
The display and generation of hard copies of the resulting
averaged image and the image intensities is supported by special
programs allowing the generation of SUN rasterfiles, a format
accepted by most image processing softwares.
OPTIONS
Filter window size and size of window used to calculated local
statistics may be selected indivicually.
SEE ALSO
typedef_ISP.h, average_filter .
DIAGNOSTICS
All messages are generally self-explanatory.
NOTES
Due to the calculation of the local statistics prior to the
filtering the program may run rather rather slow.
© Copyrights for Documentation, Users Guide and Reference Manual by Gamma Remote Sensing, 1997.
UW, CW, last change 6-Jan-1997.